nnpd.results

Small result readers: no numbered configuration IDs, pandas or hidden plotting state.

Functions

runs(root, *[, complete_only])	Runs in an output folder; by default only those with a complete analysis.
value_at(value, path)	Dotted dictionary/list lookup, e.g. coverage.projected_mean.0.
export_csv(root, destination, columns)	columns maps a CSV heading to a dotted path inside a complete run record.
comparison_plot(records, x_path, metric_path, *[, ...])	Return a figure; callers freely filter records and choose any numeric metric.

Module Contents

nnpd.results.runs(root, *, complete_only=True)

Runs in an output folder; by default only those with a complete analysis.

Each record is run.json (training settings, member, backend) plus “status” (training), “analysis” (analysis settings and state), “metrics” and “path”.

nnpd.results.value_at(value, path)

Dotted dictionary/list lookup, e.g. coverage.projected_mean.0.

nnpd.results.export_csv(root, destination, columns)

columns maps a CSV heading to a dotted path inside a complete run record.

nnpd.results.comparison_plot(records, x_path, metric_path, *, group_path='member.prior')

Return a figure; callers freely filter records and choose any numeric metric.